Machine Learned Models for Estimation of Sleep Quality in Free-Living Accelerometer Data

Background

Lists can optionally be displayed incrementally:

First item
Second item
Third item

Insert pauses to make other types of content display incrementally.

Methods

Overview Network plot of data preparation steps

network plot

Exclusion Criteria

Features

Basic Features

Weekday
Time of Day
Placement
Temperature

ACC derived features¹

Mean ACC X
Mean ACC Y
Mean ACC Z
Standard Deviation X
Standard Deviation Y
Standard Deviation Z
Max Standard Deviation
Inclination

Sensor-Independent Features²

Clock Proxy Linear
Clock Proxy Cosinus

Human Circadian Clock

Forger, Jewett, and Kronauer (1999): a so-called cubic van der Pol equation

\[\frac{dx_c}{dt}=\frac{\pi}{12}\begin{cases}\mu(x_c-\frac{4x^3}{3})-x\begin{bmatrix}(\frac{24}{0.99669\tau_x})^2+kB\end{bmatrix}\end{cases}\]

This thing is dependent on ambient light and body temperature!

Walch et al. (2019) incorporated this feature using step counts from the Apple Watch

But as demonstrated by Walch et al. (2019), a simple cosine function does the tricks just as well :)

Circadian Proxy Features

	Logistic Regression	Neural Network	Decision Tree	XGboost
Performance Metrics
Grouped by Event Prediction
In-bed Prediction
F1 Score	94.26%	95.88%	95.33%	95.79%
Accuracy	92.95%	95.10%	94.52%	95.01%
Sensitivity	97.96%	96.40%	94.68%	96.02%
Specificity	85.71%	93.24%	94.29%	93.56%
Sleep Prediction
F1 Score	93.08%	94.35%	93.90%	94.29%
Accuracy	90.87%	92.77%	92.29%	92.71%
Sensitivity	94.19%	92.59%	90.98%	92.20%
Specificity	84.65%	93.09%	94.73%	93.69%

Performance of the models to predict each class seperately, i.e., “sleep” and “in-bed”.

	Logistic Regression	Neural Network	Decision Tree	XGboost
Performance Metrics
Grouped by Event Prediction
In-Bed Awake
F1 Score	96.00%	96.38%	96.50%	96.54%
Accuracy	92.38%	93.09%	93.32%	93.39%
Sensitivity	97.56%	97.90%	98.17%	98.16%
Specificity	13.11%	19.66%	19.24%	20.51%
In-Bed Sleep
F1 Score	93.12%	94.33%	93.90%	94.29%
Accuracy	90.91%	92.74%	92.29%	92.71%
Sensitivity	94.25%	92.65%	90.98%	92.22%
Specificity	84.65%	92.90%	94.73%	93.65%

Performance of the models to predict each combined class, i.e., “sleep” + “in-bed”.

In-bed classification flow

Sleep classification flow

Forger, D. B., M. E. Jewett, and R. E. Kronauer. 1999. “A Simpler Model of the Human Circadian Pacemaker.” Journal of Biological Rhythms 14 (6): 532–37. https://doi.org/10.1177/074873099129000867.

Hirshkowitz, Max, Kaitlyn Whiton, Steven M Albert, Cathy Alessi, Oliviero Bruni, Lydia DonCarlos, Nancy Hazen, et al. 2015. “National Sleep Foundation’s Sleep Time Duration Recommendations: Methodology and Results Summary.” Sleep Health, 4.

Skotte, Jørgen, Mette Korshøj, Jesper Kristiansen, Christiana Hanisch, and Andreas Holtermann. 2014. “Detection of Physical Activity Types Using Triaxial Accelerometers.” Journal of Physical Activity and Health 11 (1): 76–84. https://doi.org/10.1123/jpah.2011-0347.

Walch, Olivia, Yitong Huang, Daniel Forger, and Cathy Goldstein. 2019. “Sleep Stage Prediction with Raw Acceleration and Photoplethysmography Heart Rate Data Derived from a Consumer Wearable Device.” Sleep 42 (12): zsz180. https://doi.org/10.1093/sleep/zsz180.